Influencing Driver Offending Behavior: Using an Integrated Deterrence-based Model

Deterrence theory is the framework traditionally used to underpin road policing practices. However, there have been several developments in deterrence theory. This study uses an integrated approach and tests four hypotheses derived from classical deterrence theory, Stafford and Warr’s reconceptualization and informal sanctioning. Regression analysis of self-reported data from a sample (N = 623) of Queensland’s drivers provided evidence that punishment avoidance, both direct and vicarious, significantly predicted offending driving behaviors. Moreover, while offending driving behaviors appeared unrelated to formal sanctioning, including certainty, severity, and celerity, they were inversely associated with informal sanctions involving shame, guilt, and concern over losing the respect of friends. This indicates that policing agencies may be able to reduce road offending by implementing interventions targeted at the informal sanctions associated with these behaviors.


Introduction
In the last 20 years, the death toll resulting from road traffic crashes remained largely unchanged.The World Health Organization (WHO, 2018) estimates that 1.3 million people die and as many as 50 million people are injured from road traffic crashes each year .These unacceptable figures raise concerns and flag the importance of the problem.The severe consequences of fatal crashes have motivated researchers to investigate the causes of crashes.The majority of these crashes involved a human factor which includes errors, misjudgment, and intentional risky driving behaviors.Risky driving behavior such as alcohol and drug violations, negligence in the use of seatbelts, speeding, running a red light and in-vehicle distraction were identified as key contributors to road injury and death (de Winter & Dodou, 2010;Redelmeier et al., 2003;Siskind et al., 2011;Sümer, 2003).The use of road policing to change driver behavior is a key strategy to reduce road crashes (Bates et al., 2012;Faulks et al., 2012).This paper will investigate an integrated deterrence-based model to influence the driving offending behavior.

Deterrence Theory
Deterrence is the behavioral response to perceptions of sanction threat (Homel, 1988;Meier & Johnson, 1977;Piquero & Pogarsky, 2002).The fundamental proposition relies on increasing the public's perceived risk of punishment and legal threats to inhibit individuals from committing offences (Jacobs, 2010).Classical deterrence theory posits that individuals are more likely to be deterred from offending in the future when formal sanctions and legal threats are severe, swift, and certain, resulting in a higher cost of committing an offence (Gibbs, 1975;Meier & Johnson, 1977).However, scholars recognize shortcomings within the classical deterrence framework and have looked for a more nuanced understanding of how deterrence may inform offender decision making (Akers & Sellers, 2004;Homel, 1988;Stafford & Warr, 1993).Stafford and Warr (1993) introduced additional concepts to overcome what they perceived as limitations in the classical model.Stafford and Warr (1993) suggested that individuals are affected by both specific and general deterrence, and thus, their offending behavior may be deterred through personal and vicarious experiences of punishment and sanctions.The reconceptualized model comprises four deterrent mechanisms that may have a conditioning effect on any given person: (1) personal punishment experience (e.g., commit an offence and get punished); (2) personal punishment avoidance (e.g., commit an offence and get away with it); (3) indirect punishment experience (e.g., knowing others who were caught and punished after committing an offence); lastly (4) indirect punishment avoidance (e.g., knowing others who committed an offence and got away with it).The uniqueness of the reconceptualized model is that it highlights the role of direct and indirect punishment avoidance to influence offending behavior, as well as the role of direct and indirect punishment to influence compliance behavior (Bates et al., 2017;Freeman & Watson, 2006;Sitren & Applegate, 2012).
Contemporary scholars recognize that formal sanctions are not applied within a social vacuum, and therefore, they incorporated elements of informal social control within the deterrence framework (Akers & Sellers, 2004;Grasmick et al., 1993).They argue that the threat of self-imposed shame, socially imposed embarrassment and losing the respect of friends and family would produce a deterrent effect.Specifically, those with stronger emotional attachment to family members and peers are likely to refrain from violating societal norms when their behavior may disappoint others and damage those valuable relationships.Previous research showed that individuals might comply to avoid feelings of shame, embarrassment, and guilt if others discover their deviance or punishment.The viability of such a notion is often discussed within the context of adolescent offending (Allen et al., 2017;Bossler, 2021;Meier & Johnson, 1977;Nagin, 2013;Nagin & Paternoster, 1994;Paternoster, 1987;Zimmerman, 2008).

Deterrence and Road Policing
In the road policing context, deterrence theory is the traditional framework used to underpin road policing strategies (Bates et al., 2012;Fleiter et al., 2013).It has been applied to reduce driving offending behaviors that are related to a higher risk of road trauma.Many road policing studies use the deterrence framework, including those examining drink driving (Freeman & Watson, 2006), drug driving (Armstrong et al., 2018), speeding (Truelove et al., 2017), and mobile phone usage while driving (Truelove et al., 2019).Effective mechanisms to deter driving offences that are based on deterrence theory include random breath tests (RBT) (Homel, 1988), random drug testing (Watling et al., 2010), speed and red light cameras (Watson et al., 2015), and mobile phone use cameras (Kaviani et al., 2020).
Punishment avoidance is perhaps the most potent construct within the reconceptualized deterrence model (Stafford & Warr, 1993).Piquero and Pogarsky (2002) suggested that punishment avoidance is associated with low perceptions of certainty of punishment and a higher propensity to offend.
Experiencing direct punishment avoidance or observing others in one's social group avoiding punishment is associated with offending in the future (Freeman & Watson, 2006;Paternoster & Piquero, 1995;Piquero & Paternoster, 1998).Although the findings are largely correlational, they draw attention to an alternative mechanism for understanding driving behavior.Evidence shows that direct punishment avoidance is a key concept in road policing and was a significant predictor of drink driving (Freeman & Watson, 2006;Szogi et al., 2017), drug driving (Armstrong et al., 2018;Watling et al., 2010), speeding (Freeman et al., 2017), and the use of Snapchat (a web-based instant messaging platform) among younger drivers operating a vehicle (Truelove et al., 2019).
Punishment avoidance may shape offender decision-making for several reasons, including the potential for offenders to learn new avoidance strategies.In their recent qualitative study, Bates and Anderson (2019) identified key themes on how young drivers experience direct punishment avoidance.For example, young drivers reported proactive actions to evade punishment by avoiding the commonly policed routes.Although this theme has not been tested among older drivers, the adaptive process may be similar.
Mixed results have been found when examining the effects of indirect punishment avoidance on driving behavior.Watling et al. (2010) identified that indirect punishment avoidance was the strongest predictor of the intention to drug and drive in the future.Similarly, indirect punishment avoidance predicted the usage of Snapchat while driving among young drivers (Truelove et al., 2019).In contrast, indirect punishment avoidance did not significantly predict speeding (Truelove et al., 2021) or drink driving (Szogi et al., 2017).
In line with Stafford and Warr's (1993) reconceptualized model, scholars have suggested that indirect punishment experiences will have an influential deterrent effect if it occurs in relation to emotional attachments such as close friends and family members (Sitren & Applegate, 2006, 2007).Inconsistent with the reconceptualized model, evidence suggests that the role of indirect punishment experiences is unclear and may even weaken the effectiveness of the deterrence process (Bates et al., 2017;Freeman & Watson, 2006;Watling et al., 2010).Specifically, it was found that indirect punishment experiences increased the intention to drink and drive (Piquero & Paternoster, 1998) and failed to predict the likelihood of future drug driving (Armstrong et al., 2018).The reason for these ambiguous findings may be that many potential offenders believe they are more proficient and effective than others in avoiding detection (Piquero & Pogarsky, 2002).Another reason is that the received punishment was not perceived as severe or did not have a significant impact (Taxman & Piquero, 1998).
Within the road safety arena, Homel (1988) developed a conceptual model which incorporated four informal sanctions, including social disapproval, internal loss, material loss, and physical loss to influence drink driving.This model was used to examine the deterrent effect on a wide range of driving offences, including drink driving (Freeman & Watson, 2006), speeding (Truelove et al., 2017), and mobile phones usage (Kaviani et al., 2020).
Some empirical evidence suggests that social sanctions were less influential in the case of recidivist drink driving (Freeman & Watson, 2006) and speeding (Truelove et al., 2017).Similarly, the threat of social disapproval and guilt did not influence compliance, at least with regard to young drivers (Poirier et al., 2018).
In contrast, there is anecdotal evidence that feelings of shame among young drivers may potentially have a mediating effect in relation to informal deterrence effect.Embarrassment, breaching the trust of others, and presenting an image of a responsible driver were key themes influencing speeding behavior (Fleiter et al., 2010).Additionally, Davey et al. (2008) considered losing friends' respect, being ashamed, and feeling guilty as important social sanctions to influence drug driving.This mediating process has not been fully explored using road offending data.
In summary, the above work suggests that the effects of direct and indirect punishment experiences may persist when classical deterrence measures (i.e., perceptions of certainty, severity, and swiftness) are controlled.It is also anticipated that avoidance experiences and informal punishment may go above and beyond the impact of "classical" deterrence measures, and this may occur through a number of causal mechanisms.Punishment avoidance may impact perceptions of certainty of apprehension, which may be a mechanism through which punishment avoidance works.Risk perceptions may also serve as an important mechanism or mediating process.Importantly, only some drivers experience direct punishment each year, but a much greater proportion of drivers avoid punishment, observe their peers avoid or experience punishment, or feel the social stigma of informal punishments.This research seeks to measure and empirically test these alternative deterrence processes.

Present Study
While there has been a significant amount of research investigating the association between deterrence theory and offending behavior (Pogarsky & Piquero, 2003;Sitren & Applegate, 2006) and driving behaviors more generally (Freeman & Watson, 2006;Poirier et al., 2018;Szogi et al., 2017;Truelove et al., 2017), few studies have investigated classical deterrence, reconceptualized deterrence model and informal deterrence mechanisms simultaneously.Theoretical insights in road policing may be gained by modeling punishment and punishment avoidance process, as well as indirect learning process, as predictors of driving offenses.This study builds upon the existing literature to address the gap of lacking an integrated deterrencebased model.
Therefore, the overarching aim of this study was to develop a more nuanced understanding of the relationship between deterrence factors, including formal and informal sanctions on self-reported driving offending behavior.Additionally, it investigated the influence of avoiding and experiencing punishment on continued offending behavior.The proposed integrated model is outlined in Figure 1.The figure clearly represents the theories incorporated in this study but not the hypotheses of this study.
Based on this proposed model, this study hypotheses that: H1: Classical deterrence variables (i.e., perceived certainty, severity of punishment, and swiftness of punishment) will have a significant negative relationship with self-reported driving offending behavior.

Participants and Procedure
Participants were recruited between the period of February 10th, 2021, and March 31st, 2021, via online social media platforms (e.g., Facebook, Instagram, and LinkedIn) and through Griffith University recruitment email broadcast.Individuals were invited to participate in an online survey regarding driving behavior and experiences.Those who were interested in participating in the study needed to be living in Queensland and have a valid car driving license, regardless of the license status (e.g., suspended) and class (e.g., provisional) to be eligible to take part in the research.Before commencing the survey, participants were provided with the digitized consent form and the information sheet describing the aims of the study and its voluntary nature.All individuals were informed that their participation was anonymous.At the completion of the survey, participants had the option of entering a prize draw for a $100 AUD Coles Gift Card.Ethics approval was obtained from Griffith University Human Research Ethics Committee prior to the commencement of the study (GU/Ref:2020/410).

Measures
The survey instrument was designed specifically for this study to collect information on the participants' perceptions of certainty, severity and swiftness of punishment and self-reported offending and risky driving behavior.The instrument also measured driving experiences relating to punishment avoidance.Additionally, the survey captured information relating to participants' family and close friends' driving experiences, as well as how participants felt when they engage in driving violations.
Socio-demographic characteristics.The self-report survey asked participants to provide their age measured in years, gender (female, male, or other), and ethnic background (Aboriginal/ Torres Strait Islander, Caucasian/European, or other).Participants were also asked to indicate the highest level of completed education and current employment status.Participants were required to report the class of their car license (Learner (L), Provisional (P1&P2), or Full/Open license (O)) and whether it was currently valid or suspended.Driving frequency was measured using exposure time as previous research has shown that it gauges exposure to different driving scenarios and environments (Chipman et al., 1993).Thus, participants indicated how frequently they drove: once a day, a few times a day, once a week, a few times a week, once a month, or a few times a month.Descriptive statistics are presented in the result section.It is important to note that the age when a driver receives his or her license is sometimes used as a valid measure of driving exposure.Tao et al. (2017) found that age correlated significantly with driving experience measured using driving frequency.Because of this and there is not a lot of variation in when Queensland drivers receive their license (80% at 17-18 years), we decided to use only age and driving frequency as measures of driving exposure to limit collinearity in the models.
Classical deterrence measures.Classical deterrence questions were adapted from previous research in the road policing context and focused on the perceptions of certainty, severity, and swiftness of punishment (Freeman & Watson, 2006).Participants were asked to respond to the extent they agreed or disagreed with each item on a 7-point scale from 1 = strongly disagree to 7 = strongly agree.
Perceived certainty of punishment was measured using five items.One question focused on their overall likelihood of being apprehended or caught by police for violating road rules "The chances of me being caught for violating road rules are high."The other questions measured levels of concern about being apprehended for specific driving violations, including drink driving, illegally using mobile phones while driving, speeding, and running through a yellow light.The perceived severity of punishment was assessed with two items.The severity items were "In general, the penalties I would receive for traffic offences would be severe" and "The penalties I would receive for traffic offences would have a significant impact on my life."The last classical deterrence construct was the perceived swiftness of punishment and was assessed with two items.The perceived swiftness items were "In general, the time between committing a driving offence and receiving a fine notice would be long" and "In general, the time between a traffic offence and going to court would be long."While there are similarities between penalties and fines, they are subtly different concepts.Penalties refers to any punishment that the driver could receive for committing a driving offence such as license loss or a monetary cost while a fine refers to a monetary cost only.
Reconceptualized deterrence measures.The four constructs used to measure the key elements within the reconceptualized deterrence model were adapted from the work by Freeman and Watson (2006) on drink driving.However, this study focused on a wider range of behaviors and therefore included the original drink driving items as well as additional items that measured speeding and breaking road rules.Additionally, there was a general item that captured the distracted behavior of the illegal use of mobile phones while driving.
Participants were asked to respond to each item using a 7-point Likert response scale ranging from 1 = strongly disagree to 7 = strongly agree.
Because indirect punishment and punishment avoidance have not been subject to the same level of study in the empirical literature, this study developed a more comprehensive set of items for these constructs as a methodological innovation.Indirect punishment experience was measured using 12 items.The perceived severity of punishment that the friends and family members of the participants received was measured.For example, "In general, the penalties that my friends received for traffic offences have had a significant impact on their lives."Additionally, there were questions regarding observing friends and family members being caught for traffic offences.The Cronbach's alpha for the 12 items measuring indirect punishment experiences was .82.Direct punishment avoidance was measured using five items that were measured on a 7-point scale from 1 = strongly disagree to 7 = strongly agree.For example, "I often drive over the speed limit without getting caught."It was identified in previous research that drivers actively avoid punishment by driving on roads where no speed cameras or police patrols were perceived to be present (Bates & Anderson, 2019).Thus, two additional items were developed to measure this active punishment avoidance.The two additional items included "I prefer driving on roads where no presence of police officers" and "I prefer driving on roads where no presence of speed cameras."The seven direct punishment avoidance items showed a reasonable reliability with a Cronbach's alpha score of .76.For indirect punishment avoidance, there were ten items in total.Each question was asked twice, first in regard to family members' experiences and then friends' experiences measured on a Likert scale ranging from 1 = strongly disagree to 7 = strongly agree.Example items include, "My family members often break the road rules without being caught" and "My friends often break the road rules without being caught."The scale had a Cronbach's alpha of .83.
Direct punishment was captured with one question, "Have you received a fine for breaking road rules?" which participants could respond with yes or no.An important distinction between this item and the classical deterrence items is that this question measures whether or not the driver has actually received a penalty.In contrast, the classical deterrence items measure perceptions of legal sanctions (i.e., how swift, certain, and severe they are perceived to be), regardless of whether or not they have ever received a penalty.While it could be assumed that experiencing a direct punishment would have the same effect as the classical deterrence perceptual measures, we have sought to test this empirically.
Informal deterrence.The survey included six questions to measure informal deterrence, based on the work of Homel (1988) regarding perceptions of social disapproval, guilt, and concerns about material loss.Homel (1988) proposed that legal sanctions such as license disqualification and losing demerit points could have a deterrent effect when they are posed as a threat to material loss.Thus, in this study, concerns over losing demerit points or losing a driving license are considered informal deterrence mechanisms.An example of the material loss items is "I would be concerned to lose demerit points if I broke the road rules." Homel (1988) also proposed that the threat of internal loss in the form of feeling guilty or ashamed of self, influences the offending behavior.An example of the internal loss variable is "I would feel ashamed of myself if the police caught me for breaking the road rules."The stigma resulting from social disapproval is another factor that was proposed by Homel (1988).An example of the social disapproval item is "If I was to break road rules, I would be concerned that I might lose my friends' respect." Participants were asked to respond on a 7-point scale ranging from 1 = strongly disagree to 7 = strongly agree.The informal deterrence scale demonstrated a reasonable level of reliability with a Cronbach's alpha score of .78.The reported Cronbach's alpha is consistent with Truelove et al. (2017) measures in which they reported a Cronbach's alpha of .92 for social sanctions items (e.g., lose friend's respect), a Cronbach's alpha of .89for internal loss items (e.g., feeling ashamed) and a Cronbach's alpha of .86 for material loss items (e.g., loss demerit points).
Self-reported driving offending behavior.The dependent variable in this study was self-reported offending driving behavior and included eight items.The Driver Behavior Questionnaire (DBQ) was used to measure offending driving behavior.The DBQ measures driving errors, lapses, and violations (Reason et al., 1990).However, it does not include any items addressing mobile phone use while driving.Previous research has indicated that lapses and driving errors have poor predictive relationship with negative driving outcomes (Lawton et al., 1997).This study focuses on the deliberate deviation from lawful driving behaviors.In this study, four items related to disregarding speed limits, drink driving and running a red light were adapted from the Driver Behavior Questionnaire (DBQ) violations subscale (Lajunen et al., 2004).The other four items were developed to capture distracted driving behaviors.Previous research explored deterrence measures in relation to the illegal use of mobile phones while driving including holding the mobile phone and conversing while driving (McCartt et al., 2010), texting (Ehsani et al., 2014) and using social media apps such as Snapchat (Truelove et al., 2021;Truelove et al., 2019).Thus, each of these behaviors is captured using one item as well as an additional item that captures the general use of the mobile phone for tasks that requires handholding.Respondents were asked to indicate how often they engage in each violation on a 6-point scale ranging from 1 = never to 6 = nearly all the time.The Cronbach's alpha for the scale was .80.

Analysis
The survey data were analyzed using the Statistical Package for the Social Sciences (SPSS) V.27.Descriptive statistics in terms of mean and standard deviation for each variable was calculated and presented to describe the data.The significance level was set at .05 unless mentioned otherwise.The sample included 25 partially completed surveys.Specifically, these surveys had 90% of the questions completed and were therefore included in the analysis.Missing values were excluded listwise from the analyses as there was little impact on the sample size.There were an additional 276 surveys that had less than 90% of the questions answered and therefore not included in the sample.These surveys were not qualitatively different than the completed ones after comparing the basic characteristics of cases with missing data and those without missing data.However, it was decided not to include these surveys to prevent nonresponse bias and avoid invalid conclusions (Bennett, 2001;Cook et al., 2000).
The various scales used in this study were assessed for reliability.Scale reliability analyses were conducted to combine the scale items into a total score (Howitt & Cramer, 2008).Cronbach's alpha was computed to measure the internal consistency of the scale.While Cronbach's alpha scores greater than .80 are considered strong scales, those between .70 and .80 are considered to have adequate reliability (Bresciani et al., 2009).The summary of Cronbach's alpha for each scale used in this study is presented in Table 1.
A hierarchical regression model was conducted to explore the influence of formal and informal deterrence and punishment avoidance compared to My friends have been caught for holding their phones while driving.

1.46
My family members have been caught for holding their phones while driving.

1.34
In general, the penalties that my friends received for traffic offences have been severe.

1.67
In general, the penalties that my family members received for traffic offences have been severe.

1.66
In general, the penalties that my friends received for traffic offences have had a significant impact on their lives.
3.17 1.70 In general, the penalties that my family members received for traffic offences have had a significant impact on their lives.
3.12 1.70 punishment experiences on self-reported driving behavior.In terms of model specifications, potential predictors of self-reported offending driving behavior were entered in blocks.Socio-demographic variables were entered at block 1 to control for their contribution towards offending driving behavior.
In the second block, the classical deterrence variable was entered to control for the influence of the threat of legal sanctions.Afterwards, the key constructs of the reconceptualized deterrence were entered in block three.Finally, the inclusion of informal deterrence occurred in the final block.

Participants
The sample size for this study was 623 participants.All participants were from Queensland, Australia and held a car driver's license.Of these, 84.1% (n = 524) were recruited through social media platforms, and 15.9% (n = 90) were recruited via the university-wide invitation email.
As shown in Table 2, 43.8% of the sample were females (n = 273), and 55.1% were males (n = 343).Participants were categorized according to their age into three groups, younger drivers (16-25 years old), middle-aged drivers (26-55 years old) and older drivers (>56 years old).The categorization approach was consistent with previous research on driving behavior (Vardaki & Yannis, 2013;Zhang et al., 2013).Participants were aged between 16 and 80 years old, with an average age of 42.06 (SD = 18.91).Further, most of the

Prediction of Offending Behavior Using Three Types of Deterrence
Table 3 presents the results from the hierarchical regression conducted to explore the influence of the three types of deterrence in predicting offending driving behavior and the influence of punishment avoidance compared to punishment experiences in predicting self-reported offending driving behavior.The results of the first block show that the socio-demographic variables contributed significantly to the model, F (8,589) = 13.08,p < .001with an adjusted R 2 of 0.14, explaining 14% of the variation in self-reported offending driving behavior.Age was a significant predictor β = −.23,p < .001,indicating that young drivers were more likely to report offending driving behavior.Further, driving frequency was a significant predictor of selfreporting offending driving behavior β = .07,p < .05,indicating that those who drove more frequently were more likely to report offending driving behavior.
In block 2, the classical deterrence variable was entered.This model was statistically significant, F (9, 588) = 12.09, p< .001with an adjusted R 2 of 0.14, explaining 14% of the variation in self-reported offending driving behavior when controlling for socio-demographic variables.The classical deterrence mechanism was not able to predict self-reported offending driving behavior.It explained a minimal additional 1% of the variance, R 2 change = 0.01.
Model 3, including the reconceptualized deterrence constructs, was also statistically significant F (13, 584) = 28.85,p < .001with an adjusted R 2 of 0.38, accounting for 38% of the variance in self-reported offending driving behavior.Direct punishment avoidance was a significant predictor of selfreporting offending driving behavior, β = .44,p < .001.Similarly, indirect punishment avoidance was a significant predictor of self-reporting offending driving behavior β = .08,p < .05.These results show that the more participants experience punishment avoidance, the more likely they were to selfreport offending driving behavior.Additionally, those who have a friend or a family member who breaks the road rules while driving without being punished were more likely to report offending driving behaviors.The reconceptualized constructs were able to explain an additional 23% of the variance, R 2 change = 0.23.The informal deterrence variable was entered in block 4. Model 4 was statistically significant, F (14,583) = 28.01,p < .001with an adjusted R 2 of .39.The model results show that informal deterrence significantly predicted self-reported offending driving behavior, β = −.12,p < .001.The results suggest that those who are threatened by internal, or material loss are more likely to report lower levels of offending driving behavior.Informal deterrence was able to explain an additional 1% of the variance.

Discussion
The current study extended the focus beyond classical deterrence constructs in predicting offending driving behavior by considering alternative forms of deterrence.More precisely, this study identified that informal deterrence was a significant predictor of self-reported offending driving behavior.In addition, the hierarchical regression model showed that informal deterrence had a negative influence on offending driving behavior.These results suggest that social sanctions, the feeling of shame and embarrassment, and the fear of losing demerit points or losing a driving license were predictive of reduced offending behavior.These results are consistent with previous research on drink driving (Baum, 1999;Grasmick et al., 1993).In contrast to these findings, Truelove et al. (2017) found that social sanctions and internal loss were not predictors of speeding.The reason for this could be that drivers do not perceive different aberrant behaviors at equal significance.
In the current study, classical deterrence was not a significant predictor of self-reported offending driving behavior.To demonstrate, perceptions of certainty of punishment did not reach significance.This is because the current study suggested a relatively poor perception of certainty for offending driving behavior.Consistent with previous research, drivers do not perceive the likelihood of apprehension for breaking the road rules to be high (Fleiter et al., 2009;Szogi et al., 2017).
Besides increasing road policing activities and presence, there was an early suggestion that mass media campaigns play an important role in increasing public perceptions of the certainty of punishment (Homel, 1988).In their review Davey and Freeman (2011) concluded that well-run media campaigns can enhance the deterrent threat of apprehension and legal sanctions.However, Guttman (2015) suggests that a key challenge associated with the use of media campaigns based on deterrence principles such as the threat of being caught and the severity of penalties is how to depict policing of these activities in a positive way.He suggests that further research is required.
Similarly, perception of the severity of punishment did not reach significance.This could be because the severity of punishment does not influence individuals who do not perceive the punishment to be certain (Grasmick & Bryjak, 1980).However, the positive relationship between perceptions of certainty and severity of punishment supported the notion that high perceptions of severity are likely to follow if there are high perceptions of certainty (Homel, 1988).
Perception of the swiftness of punishment also did not reach significance.Consistent with previous research, perceived swiftness of punishment failed to deter drivers in the case of drug driving (Davey et al., 2008), drink driving (Szogi et al., 2017), and speeding (Truelove et al., 2017).In addition, the extended timeframe between getting caught and appearing in court for drink driving and drug driving could delay the consequences of illegal behavior and thus, weaken the swiftness effect (Nagin & Pogarsky, 2001).Similarly, speeding and being detected by a speed camera takes time until the offender receives the fine or feels its financial effects.In this study, perceptions of the swiftness of punishment were measured using two items to address both delayed receiving a fine notice or going to court.However, participants' responses for both items show that they somewhat disagree that the time between committing a driving offence and receiving the fine notice or going to court would be long.
While punishment experiences did not predict self-reported offending driving behavior, punishment avoidance experiences predicted self-reported offending driving behavior.More precisely, direct punishment avoidance was the strongest predictor of offending driving behavior.This is consistent with previous research, which found that direct punishment avoidance was the strongest predictor of various driving offending behavior, including drink driving (Freeman & Watson, 2006), drug driving (Armstrong et al., 2018) and using Snapchat while driving (Truelove et al., 2019).This result was consistent with a previous study suggesting that experiencing direct punishment avoidance might reduce levels of perceived certainty of punishment (Piquero & Pogarsky, 2002).Furthermore, indirect punishment avoidance was also a significant predictor.However, it was less influential than direct punishment avoidance.The results indicate that individuals are influenced by knowing a family member or friend who has not been caught speeding, running a red light, drink driving, or using their mobile phones while driving.
Direct punishment experiences and indirect punishment experiences were not significant predictors of self-reported offending driving behavior.Although these findings contradict Stafford and Warr (1993) notion of the deterrent effect of punishment experiences, they are consistent with previous research (Armstrong et al., 2018;Freeman & Watson, 2006).There was a significant positive relationship between direct punishment experiences and self-reported offending driving behavior.This relationship can be seen as an "emboldening effect" in which direct punishment experiences might encourage offending rather than deterring it (Allen et al., 2017).Piquero and Pogarsky (2002) offer an explanation for this phenomenon.They argued that defiance accounts for the emboldening effect and previous punishment experiences might be seen as unjust and as result, it triggers a form of social protest.Self-serving bias could be another explanation for the emboldening effect.Potential offenders might perceive that they became immune after receiving the punishment and they can avoid detection in the next offending episode (Bates et al., 2017).Moreover, it might be the case that the item used to measure direct punishment experience was inadequate in capturing the levels of punishment certainty and severity.
Age emerged as a significant predictor of offending driving behavior.The finding that self-reported offending driving behavior was more robust among young drivers is consistent with previous studies indicating that younger drivers are more problematic drivers and less compliant with road rules than middle-aged drivers and older drivers (Armstrong et al., 2018;Bates et al., 2017).Moreover, younger drivers reported lower direct punishment experiences suggesting that deterrence-based approaches may be less effective or, alternatively, they have not had ample time to offend and get caught yet.While Chapman et al. (2014) found that the peak traffic violation rates for young drivers happened after they turn 18 (i.e., having a full license), the former suggestion might provide some insights.
Further, younger drivers reported higher direct punishment avoidance experiences than middle-aged or older drivers.Bates and Anderson (2019) had three explanations for this phenomenon.Firstly, they suggested that parents might take responsibility for the driving offences and facilitate punishment avoidance, limiting the deterrent effect.This can explain why, in this study, they did not perceive the punishment to be severe.Secondly, younger drivers actively engage in punishment avoidance behaviors by avoiding roads with police presence or removing their P-plates that indicate their early driving status.Thirdly, many younger drivers avoided punishment by persuading the police officer not to issue an infringement notice.
Similar to previous research, driving frequency was a significant predictor of engaging in driving offending behavior, indicating that high exposure to roads is associated with driving offending behavior (Truelove et al., 2017).In this study, gender was not a significant predictor of self-reported driving offending behavior.This contrasts with the findings from previous studies that showed that males are more likely to engage in driving offending behavior because they discount negative future consequences (Allen et al., 2017;Freeman et al., 2017).

Implications for Road Safety
This research provided important implications for deterrence theory.Firstly, the combined deterrence model including classical deterrence constructs, the reconceptualized deterrence model and informal deterrence explained 40% of the variance in self-reported offending behavior.
Secondly, the failure of legal sanctions to predict compliance suggests that high levels of police resources and commitment could boost the effect of both specific and general deterrence.Highly visible enforcement operations at random locations and continuous efforts may increase levels of perceived certainty of detection.While deterrence requires high levels of police resources and commitment in order to maintain its effect and reduce punishment avoidance experiences, the implementation of non-legal countermeasures that affect the moral factor and trigger the internal feelings of guilt, shame and embarrassment over engaging in offending behavior could influence the road offending behavior.One possible method of implementing this could be, when police officers are issuing a traffic ticket in person, engaging in a discussion with the driver regarding how a family member, such as a parent or spouse, would feel about them being punished for breaking a road rule.This discussion may explicitly trigger internal feelings of guilt, shame and embarrassment that could lead to a change in behavior.
Moreover, law enforcement agencies and policy makers need to look beyond punishment to identify other methods to increase awareness and encourage compliance.One method that has been effective in reducing drink driving crashes is mass media campaigns (Davey & Freeman, 2011).Well executed and well implemented mass media campaigns that focus on personal and social costs associated with road offending could have the potential to reduce offending driving behavior (Elder et al., 2004).

Strengths and Limitations
Consistent with previous research within the deterrence and road safety field, this study is based on self-reported data, which may be vulnerable to selfreport bias.It should be borne in mind that participants might have difficulties in remembering details of punishment experiences and punishment avoidance experiences.However, self-reported data related to offending driving behaviors are considered to have an acceptable level of validity when collected anonymously, and there are no consequences or prosecutions associated with responding (Lajunen & Summala, 2003).Objective measures using in-vehicle monitoring systems would be beneficial for future research instead of relying on self-reported data.
The cross-sectional data used here is not without limitations and the results should be interpreted within this context.The lack of longitudinal data deprived us form addressing the temporal ordering of variables in terms of the offending driving behaviors and the outcome of sanctioning experiences and how the formulation of perceptions changes over time.However, previous research investigated deterrence in the road safety domain measured offending behavior on a scale ranging from "never" to "all the time" (Ochenasek et al., 2022;Szogi et al., 2017;Truelove et al., 2017).Despite this, we acknowledge that the cross-sectional data used in this study neglected the number of prior experiences, the nature, the timing and the recency of past behaviors.To address this limitation in the future, it might be beneficial to replicate the most recent work done in the road safety space by Truelove et al. (2020).In their study, they examined the stability of perceptions of classical and informal deterrence variables over time among young drivers aged between 17 and 25 years old.They concluded that deterrence is a dynamic process and perceptions fluctuate with time for various reasons such as less driving experience, the lack of exposure to countermeasures, punishment avoidance and prior offending experiences.
There were low mean scores for some of the items, making it difficult to accurately picture drivers' behavior and their perceptions of legal and nonlegal sanctions.For example, the mean score for driving over the legal BAC limit was M = 1.22, which was relatively low.To obtain precise and accurate knowledge, future researchers should capture how many times drivers engage in such behaviors over a specific period of time.Moreover, the Cronbach alpha reliability score for classical deterrence scale was low (α = .59)which raises a question concerning the validity of the classical deterrence items and scale.It might be beneficial in the future to use more items to develop a reliable classical deterrence scale or use independent scales for each of the three classical deterrence constructs.Future deterrence-based research could be extended to include additional variables such as personality traits, perceptions of risk and perceptions of safety to encourage compliance with road rules.
This study used a convenience sample of Queensland's drivers.However, the demographic breakdown of the sample differed slightly from the population of Queensland.According to Queensland census data in 2021, the mean

Table 1 .
Descriptive Statistics and Cronbach's Alpha for the Scales.

Table 2 .
Descriptive Statistics Summary for Socio-Demographic Characteristics.